Daily snapshot · immutable permalink
AI model usage rankings - September 23, 2026
As published on September 23, 2026, DeepSeek: DeepSeek V4.1 Flash (batch) led real OpenRouter usage share at 14.2%. Space Bunny Alpha, Anthropic: Claude Opus 5.5 (batch), OpenAI: GPT-6 Sol (batch), OpenAI: GPT-6 Luna Pro (batch), Anthropic: Claude Opus 4.8 (batch) entered the leaderboard for the first time. OpenAI: GPT-6 Luna (batch) moved 27 places up since September 22, 2026.
- #1 DeepSeek: DeepSeek V4.1 Flash (batch) - 14.2% usage share
- #2 Z.ai: GLM 5.3 Flash (batch) - 12.1% usage share
- #3 DeepSeek: DeepSeek V4 Flash 0423 - 8.6% usage share
- #4 Tencent: Hy4 preview - 8.2% usage share
- #5 OpenAI: GPT-5.6 Luna (batch) - 6.3% usage share
- #6 Xiaomi: MiMo-V2.6-Flash - 2.9% usage share
- #7 NVIDIA: Nemotron 3 Ultra (free) - 2.4% usage share
- #8 Tencent: Hy3 - 2.3% usage share
- #9 Z.ai: GLM 5.3 (batch) - 2.3% usage share
- #10 OpenAI: GPT-6 Luna (batch) - 2.1% usage share
| # | Model | Usage share | Rank change |
|---|---|---|---|
| 01 | DeepSeek: DeepSeek V4.1 Flash (batch) | 14.2% | +1 (+0.9pts) |
| 02 | Z.ai: GLM 5.3 Flash (batch) | 12.1% | -1 (-3.3pts) |
| 03 | DeepSeek: DeepSeek V4 Flash 0423 | 8.6% | +1 (0pts) |
| 04 | Tencent: Hy4 preview | 8.2% | -1 (-2pts) |
| 05 | OpenAI: GPT-5.6 Luna (batch) | 6.3% | no change |
| 06 | Xiaomi: MiMo-V2.6-Flash | 2.9% | +17 (+2.1pts) |
| 07 | NVIDIA: Nemotron 3 Ultra (free) | 2.4% | -1 (-1.2pts) |
| 08 | Tencent: Hy3 | 2.3% | no change |
| 09 | Z.ai: GLM 5.3 (batch) | 2.3% | +1 (+0.6pts) |
| 10 | OpenAI: GPT-6 Luna (batch) | 2.1% | +27 (+1.7pts) |
| 11 | OpenAI: GPT-5.6 Sol (batch) | 1.9% | +4 (+0.7pts) |
| 12 | Xiaomi: MiMo-V2.5 | 1.7% | -5 (-1.7pts) |
| 13 | Google: Gemini 3.8 Flash (batch) | 1.6% | -4 (-0.3pts) |
| 14 | Space Bunny Alpha New | 1.5% | - |
| 15 | DeepSeek: DeepSeek V4 Pro 0423 | 1.4% | -4 (0pts) |
| 16 | Z.ai: GLM 5.2 | 1.3% | -2 (+0.1pts) |
| 17 | Meta: Muse Spark 1.3 Contributor | 1.3% | -4 (0pts) |
| 18 | Anthropic: Claude Sonnet 5 (batch) | 1.3% | -2 (+0.1pts) |
| 19 | Upstage: Solar Pro 4 | 1.2% | -7 (-0.1pts) |
| 20 | MiniMax: MiniMax M3 | 1.1% | -2 (+0.1pts) |
| 21 | Anthropic: Claude Opus 5.5 (batch) New | 1.0% | - |
| 22 | MoonshotAI: Kimi K3 (batch) | 1.0% | -3 (0pts) |
| 23 | inclusionAI: Ling 3.0 Flash Fin | 0.9% | -2 (+0.1pts) |
| 24 | OpenAI: GPT-6 Astra (batch) | 0.9% | no change |
| 25 | Xiaomi: MiMo-V2.6-Pro | 0.8% | -5 (-0.1pts) |
| 26 | Poolside: Laguna S 2.1 (free) | 0.8% | -4 (0pts) |
| 27 | Anthropic: Claude Opus 5 (batch) | 0.8% | -10 (-0.3pts) |
| 28 | OpenAI: GPT-6 Sol (batch) New | 0.7% | - |
| 29 | Google: Gemini 3 Flash Preview (batch) | 0.6% | -4 (0pts) |
| 30 | Meta: Muse Spark 1.3 | 0.5% | -1 (+0.1pts) |
| 31 | Google: Gemini 2.5 Flash Lite (batch) | 0.5% | -3 (0pts) |
| 32 | OpenAI: gpt-oss-120b (batch) | 0.5% | -6 (0pts) |
| 33 | OpenAI: GPT-5.6 Terra (batch) | 0.4% | -6 (0pts) |
| 34 | Google: Gemini 3.1 Flash Lite (batch) | 0.4% | -1 (0pts) |
| 35 | Google: Gemini 2.5 Flash (batch) | 0.4% | -1 (0pts) |
| 36 | Qwen: Qwen3.8 Flash | 0.4% | -5 (0pts) |
| 37 | Google: Gemini 3.7 Flash (batch) | 0.4% | -5 (0pts) |
| 38 | Qwen: Qwen3.7 Flash | 0.3% | +5 (+0.1pts) |
| 39 | NVIDIA: Nemotron 3.5 Lightning (free) | 0.3% | no change |
| 40 | Qwen: Qwen3.8 27B | 0.3% | +4 (0pts) |
| 41 | inclusionAI: Ling 3.0 Flash VL | 0.3% | -11 (-0.1pts) |
| 42 | Google: Gemma 4 31B (free) | 0.3% | -2 (0pts) |
| 43 | Z.ai: GLM 5.3 FlashX | 0.3% | -5 (-0.1pts) |
| 44 | Anthropic: Claude Fable 5.1 (batch) | 0.3% | -9 (-0.1pts) |
| 45 | OpenAI: GPT-6 Luna Pro (batch) New | 0.3% | - |
| 46 | Anthropic: Claude Opus 4.8 (batch) New | 0.3% | - |
This page is a permanent, unchanging record of real OpenRouter usage share as it stood on September 23, 2026. Later corrections to the data pipeline appear as new dated snapshots - this page's figures never change.
Source: OpenRouter (openrouter.ai/rankings), as of September 23, 2026.
Benchmark scores: Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings).
Methodology: www.smophy.ai/benchmark/methodology
Token counts originate from each provider's own tokenizer and are not directly comparable across providers.
